海量三维点云数据的组织与可视化研究

发布时间：2018-01-28 01:44

本文关键词： 地理场景三维点云空间数据组织海量数据空间索引空间数据可视化　出处：《南京师范大学》2013年硕士论文　论文类型：学位论文

【摘要】：随着数字城市的建设,大规模三维数据采集技术的迅猛发展,三维激光扫描,航天/航空影像的密集匹配等产生了海量点云数据,最直接的体现就是点的密度越来越大,点的数量越来越多,其中利用机载/车载/地面激光扫描系统所获取的数据,可达几十甚至上百G。现有的三维点云处理软件,如FARO公司的Focus3D点云扫描处理软件SCENE5.0、Leica公司的HDS三维激光扫描仪配套的Cyclone6.0以及Polywork、Geomagic等软件,各有所长,但是还是侧重于解决处理建模方面的问题。对海量点云支持方面较差,其原因是对点云数据组织与调度非优化方式。针对存在的问题,本文在对地理场景三维点云数据获取研究的基础之上,着重研究了海量三维点云数据组织与空间索引,分析了当前常用的三维点云数据的空间索引方法,提出改进八叉树的三维点云数据的组织与索引,并利用混合索引的方法,以降低内存的消耗以及提高查询的效率。在此基础之上,综合利用内存文件映射、可见性判别以及多层次LOD技术,降低点云绘制的数目,可在普通PC机上达到快速、高效的点云绘制。在理论和方法研究的基础之上,开发了海量三维点云可视化原型系统,验证本文提出的算法有效性。主要成果可总结如下： (1)研究了地理场景中三维点云数据的获取,主要有机载/车载/地面三维激光扫描以及航天航空/地面立体摄影影像匹配技术,对不同的获取方法进行相应的评价,总结了所获取的点云文件格式,将其统一转化为本文需要的二进制文件形式。 (2)研究了海量点云数据的组织与空间索引方法,分析了常用的三维点云数据的索引方法并对其进行总结评价,提出了改进八叉树的编码方案,在此基础上进一步提出了对叶节点数据采用KD树进行混合索引,降低了内存的消耗并提高了检索的效率。 (3)在对海量三维点云数据组织与空间索引基础之上,在点云可视化时,提出了综合运用内存文件映射、可见性判断以及多层次LOD等技术,降低在点云绘制时点云的数量,采取这些优化调度的方式,可在普通PC机上实现对海量点云的可视化。 (4)为验证本文提出算法,开发了海量三维点云可视化原型系统,验证了本文方法的有效性。
[Abstract]:With the construction of digital city and the rapid development of large-scale 3D data acquisition technology, 3D laser scanning, space / aviation image dense matching has produced massive point cloud data. The most direct manifestation is that the density of points is increasing and the number of points is increasing. Among them, the data obtained by airborne / vehicle / ground laser scanning system is more and more. Existing 3D point cloud processing software, such as FARO's Focus3D point cloud scanning software SCENE5.0. The HDS 3D laser scanner of Leica company has its own strong points, such as Cyclone6.0 and Polywork Geomagic. However, it is still focused on solving the problem of modeling. It is poor in support of massive point cloud because of the non-optimization of point cloud data organization and scheduling. Aiming at the existing problems, this paper focuses on the organization and spatial index of massive 3D point cloud data based on the research of 3D point cloud data acquisition in geographic scene. The spatial index method of 3D point cloud data is analyzed, and the organization and index of 3D point cloud data based on octree are improved, and the mixed index method is used. In order to reduce the memory consumption and improve the efficiency of query. On the basis of this, the use of memory file mapping, visibility discrimination and multi-level LOD technology to reduce the number of point cloud rendering. On the basis of theoretical and methodological research, a massive 3D point cloud visualization prototype system is developed. The main results can be summarized as follows: 1) the acquisition of 3D point cloud data in geographic scene is studied, including airborne / vehicle / ground 3D laser scanning and aerospace / ground stereo image matching technology. The different acquisition methods are evaluated and the obtained point cloud file format is summarized and transformed into the binary file form which is needed in this paper. Secondly, the organization and spatial index method of massive point cloud data are studied, and the commonly used indexing methods of 3D point cloud data are analyzed and evaluated, and an improved octree coding scheme is proposed. On this basis, the mixed index of leaf node data using KD tree is put forward, which reduces the memory consumption and improves the efficiency of retrieval. On the basis of organizing massive 3D point cloud data and spatial index, this paper puts forward some techniques such as memory file mapping, visibility judgment and multi-level LOD in point cloud visualization. By reducing the number of point clouds when the point clouds are drawn, and adopting these optimal scheduling methods, the visualization of mass point clouds can be realized on ordinary PC computers. In order to verify the proposed algorithm, a massive 3D point cloud visualization prototype system is developed, which verifies the effectiveness of the proposed method.
【学位授予单位】：南京师范大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：P208

【参考文献】